Re: asm disks

From: Hans Forbrich <fuzzy.graybeard_at_gmail.com>
Date: Mon, 04 Feb 2013 15:27:27 -0700
Message-ID: <511035CF.8030707_at_gmail.com>



On 04/02/2013 1:37 PM, Mike Hayes wrote:
> It seems to me we have just gone against best practice. For those who have
> experience with ASM do you use hardware raid or not?
>

The 'common wisdom' is to use RAID 1+0 or RAID 5 and leave ASM at External Redundancy.

Alex Gorbachev has an interesting presentation and demo in which he demonstrates that RAID under Oracle can end up with data loss, if a second or third disk goes out. Never say never - in a typical SAN, the mean time betwen failure of any disk on the SAN is in the 'days' range, and it is conceivable that a second disk supporting the database can fail before the first is replaced and rebuilt.

On the other hand, ASM normal or high redundancy will stop the database cold, raising an error and killing the DB processes, thereby guaranteeing that there is no data loss. Transactions are either committed, and therefore in the redundant redo logs, or not committed and therefore thrown away.

His basic presentation can be seen at
http://www.nyoug.org/Presentations/2012/June/Gorbachev_Oracle_ASM.pdf - but it really only hits home with the demo in which he kills several disks at random (audience calls out which to kill) to prove the point.

So ASM does best with JBOD, normal or high redundancy and no RAID.

But of course, the common wisdom says use RAID 5 ... or if pushed (and totally against the recommendation of the SAN administrator) perhaps 1+0.

/Hans

--
http://www.freelists.org/webpage/oracle-l
Received on Mon Feb 04 2013 - 23:27:27 CET

Original text of this message